Forests cover tens of thousands of acres in eastern national parks, providing habitat for countless plants, animals, fungi, and insects. Beyond offering beautiful landscapes for outdoor recreation and learning about our nation’s history, park forests protect water quality and soil stability, as well as influence our local weather and reduce some gases that contribute to climate change.
However, these critical park resources face a range of interacting stressors: invasive plant dominance, over-abundant white-tailed deer populations, development of surrounding lands, novel pests and pathogens, altered disturbance regimes, increasingly frequent extreme weather events, and changing climate conditions.
Forest health monitoring data from five Inventory and Monitoring (I&M) Networks within NPS Region 1 have identified concerning long-term, region-wide trends in tree regeneration that indicate significant threats to future sustainability of park forests. An abundant regeneration layer containing tree seedlings and saplings (small trees) of varying sizes and composed of species found in the canopy is an important component of a healthy forest. Many parks lack the minimum level of seedling and sapling density needed to replace canopy trees as they die, whether from age, natural disturbance, or the effects of non-native pests and pathogens. When forests lack sufficient regeneration, or the species composition in the regeneration layer does not match the canopy, we call this a regeneration debt (Miller & McGill 2019). The most severe form of regeneration debt is a complete lack of regeneration, which, if allowed to persist, can ultimately lead to loss in forest habitat.
Even in parks where I&M has documented “healthy” regeneration (i.e., seedling and sapling density at levels sufficient to sustain a future tree canopy), some regeneration is composed of primarily non-canopy tree species or species whose future is known to be threatened (e.g., ash trees (Fraxinus spp.), which are being decimated by an exotic insect pest, Emerald Ash Borer). Compositional mismatch between the species in the regeneration layer and the canopy is another type of regeneration debt that can be problematic if the species in the regeneration layer are invasive, like tree-of-heaven (Ailanthus altissima), or suboptimal, like ash.
The regeneration failure observed across the region is occurring with a simultaneous increase in non-native plant abundance (Miller et al. 2020), which may further suppress regeneration. The increasing trend in non-native plants has been documented in most parks, even those where deer density and browse have been brought under control through years of deer management.
At worst, these stressors cause the parks’ native forests to be replaced by thickets of exotic invasive shrubs that impede cultural viewsheds and increase visitors’ exposure to disease-carrying ticks. Collectively, these impacts can result in increased operational costs for parks to ensure visitor safety and preserve park infrastructure and have the potential to change the character of a park’s specific resource that visitors have come to enjoy. Managing for resilient park forests is imperative to ensure long-term ecosystem health, maintain biodiversity, and meet the NPS mission of preserving park resources unimpaired for future generations.
This document summarizes status and trends in forest structure and diversity, primarily related to tree regeneration, using data collected by the National Park Service Inventory and Monitoring Division (I&M) in 39 parks spanning five I&M networks in Region 1 (Figure 1). Trends are based on change over time across three complete survey cycles, with Cycle 1 spanning 2008 – 2011, Cycle 2 spanning 2012 – 2015, and Cycle 3 spanning 2016 – 2019. Status is based on the most recent 4-year survey period that is consistent among all parks, which is 2016 – 2019. These results are currently in draft and are being used to frame multiple reports and manuscripts on forest regeneration in eastern parks. This document is not to be shared publicly.
Figure 1. Map of parks included in regional regeneration project.
This study analyzed data from 1515 plots in 39 national parks that have similar methods for long-term forest monitoring (Figure 1). In each park, monitoring plot locations were determined using Generalized Random-Tessellation Stratification (GRTS) to generate a spatially balanced and randomized sample of plot locations across the park’s forested area (Stevens and Olsen 2004). Plots were sampled on a 4-year rotating panel, such that one quarter (i.e., one panel) of the plots is sampled every year, and each plot is sampled every 4 years (i.e., one cycle).
We calculated plot-level metrics of forest structure and diversity to assess status and trends in forest health, with species ephasis on regeneration. To estimate trends over time, we used cycle as a numeric independent variable in our models, with cycle 1 covering survey years 2008 - 2011, cycle 2 covering survey years 2012 - 2015, and cycle 3 covering survey years 2016 - 2019. We fit linear mixed effects models, with plot as a random intercept, to estimate trends in forest metrics using the lme4 package (Bates et al. 2015). Diagnostics (e.g., residual plots) on these models consistently indicated issues with normality and constant variance. While estimates of coefficients (e.g., slope) are robust to violations of non-normal error, traditional significance testing is not (Maas and Hox 2004, Givens and Hoeting 2012). We therefore used case bootstrapping, a nonparametric bootstrap method, to generate empirical 95% confidence intervals of model coefficients based on 1000 samples for each model. Case bootstrapping works by randomly sampling plots (i.e., cases) along with the survey data from those plots in the order they were sampled to generate a sampling distribution that maintains the underlying random structure of the dataset (Givens and Hoeting 2012). While case bootstrapping relaxes the assumptions of the underlying error distribution, it also requires a sufficient number of plots to sample because the sampling distribution is derived entirely from resampling the existing data. We therefore were unable to assess Sagamore Hill NHS (SAHI) and Wolf Trap Park for the Performing Arts (WOTR) for trends because they had too few plots (i.e., < 6 plots) to create a usable sampling distribution.
Bates, D., M. Maechler, and B. Bolker. 2015. Fitting linear mixed-effects models using lme4. Journal of Statistical Software 67:1–48.
Givens, G. H., and J. A. Hoeting. 2012. Computational statistics. John Wiley & Sons, Hoboken, New Jersey, USA.
Leak, W. B., M. Yamasaki, and R. Holleran 2014. Silvicultural guide for northern hardwoods in the northeast. Gen. Tech. Rep. NRS-132. Newtown Square, PA: U.S. Department of Agriculture, Forest Service, Northern Research Station. 46 p.
Maas, C.J., and J.J. Hox. 2004. The influence of violations of assumptions on multilevel parameter estimates and their standard errors. Computational Statistics & Data Analysis 46:427–440.
Marquis.D.A., editor. 1994. Quantitative silviculture for hardwood forests of the Alleghenies. US Forest Service. Northeastern Forest Experiment Station. General Technical Report NE-183.
Miller, K.M. and B.J. McGill. 2019. Compounding human stressors cause major regeneration debt in over half of eastern US forests. Journal of Applied Ecology 56(6): 1355-1366.
Miller, K.M., B.J. McGill, A.S. Weed, C.E. Seirup, J.A. Comiskey, E.R. Matthews, S.J. Perles, and J.P. Schmit. 2021. Long‐term trends indicate that invasive plants are pervasive and increasing in eastern national parks. Ecological Applications 31(2):p.e02239.
McWilliams, W.H., Westfall, J.A., Brose, P.H, Dey, D.C., Hatfield, M., Johnson, K., Laustsen, K.M., Lehman, S.L., Morin, R.S., Nelson, M.D., Ristau, T.E., Royo, A.A., Stout, S.L., Willard, T., Woodall, C.W. 2015. A regeneration indicator for Forest Inventory and Analysis: history, sampling, estimation, analytics, and potential use in the midwest and northeast United States. Gen. Tech. Rep. NRS-148. Newtown Square, PA: U.S. Department of Agriculture, Forest Service, Northern Research Station. 74 p. https://doi.org/10.2737/NRS-GTR-148.
Niewinski, A.T., T.W. Bowersox, and R.L. Laughlin. 2006. Vegetation Status in Selected Woodlots at Gettysburg National Military Park Pre and Post White-tailed Deer Management. Technical Report NPS/NER/NRTR–2006/037. National Park Service. Philadelphia, PA.
Stevens, D.L., and A.R. Olsen. 2004. Spatially balanced sampling of natural resources. Journal of the American Statistical Association 99:262–278.
Vickers, L.A., W.H. McWilliams, B.O. Knapp, A.W. D’Amato, M.R. Saunders, S.R. Shifley, J.M. Kabrick, D.C. Dey, D.R. Larsen, and J.A. Westfall. 2019. Using a tree seedling mortality budget as an indicator of landscape-scale forest regeneration security. Ecological Indicators 96:718-727.
Figure 2. Summary of status and trends for each metric and Macrogroup combination ordered from high to low latitude
Status metrics (row 1) are based on the most recent 4 years of data (2016:2019). Status thresholds were defined as:
Trend metrics (rows 2-5) are grouped by species type, with Total including all tree species; Native Canopy only including native, canopy forming species; Other Native including native species that are typically understory trees; and Exotic including non-native tree species. Note that ash species (Fraxinus spp.) are not considered native canopy-forming species due to impacts from Emerald Ash Borer. Trends were assessed using non-parametric bootstrapping and random intercept linear models. Only metrics with > 6 plots and with > 10% non-zero values were modeled.
Figure 3. Deer browse impacts by cycle. Bars represent proportion of plots experiencing low, medium, high and very high impacts.
Figure 4. Trends in regeneration stocking index by Macrogroup and cycle. The stocking index only includes native canopy-forming species and is an index of whether the regeneration layer is sufficient to stock the future canopy. Note that ash species (Fraxinus spp.) are not included in the stocking index. Cycle 1 = 2008:2011; cycle 2 = 2012:2015; cycle 3 = 2016:2019. Macrogroups are sorted from high to low latitude.
Figure 5. Trends in total live tree basal area (BA; sq.m/ha) by Macrogroup and cycle. Trees are stems that are >= 10 cm DBH. Cycle 1 = 2008:2011; cycle 2 = 2012:2015; cycle 3 = 2016:2019. Macrogroups are sorted from high to low latitude.
Figure 6. Trends in total live tree density (stems/ha) by Macrogroup and cycle. Trees are stems that are >= 10 cm DBH. Cycle 1 = 2008:2011; cycle 2 = 2012:2015; cycle 3 = 2016:2019. Macrogroups are sorted from high to low latitude.
Figure 7. Trends in total sapling basal area (BA; sq.m/ha) by Macrogroup and cycle. Saplings are stems that are >= 1 cm and < 10 cm DBH. Cycle 1 = 2008:2011; cycle 2 = 2012:2015; cycle 3 = 2016:2019. Macrogroups are sorted from high to low latitude.
Figure 8. Trends in total sapling density (stems/sq.m) by Macrogroup and cycle. Saplings are stems that are >= 1 cm and < 10 cm DBH. Cycle 1 = 2008:2011; cycle 2 = 2012:2015; cycle 3 = 2016:2019. Macrogroups are sorted from high to low latitude.
Figure 9. Trends in total seedling density (stems/sq.m) by Macrogroup and cycle. Seedlings are stems that are >= 15 cm tall and < 1 cm DBH. Cycle 1 = 2008:2011; cycle 2 = 2012:2015; cycle 3 = 2016:2019. Macrogroups are sorted from high to low latitude.
Figure 10. Trends in live tree basal area by species group (BA; sq.m/ha) by Macrogroup and cycle. Trees are stems that are >= 10 cm DBH. Cycle 1 = 2008:2011; cycle 2 = 2012:2015; cycle 3 = 2016:2019. Macrogroups are sorted from high to low latitude.
Figure 11. Trends in live tree density by species group (stems/ha) by Macrogroup and cycle. Trees are stems that are >= 10 cm DBH. Cycle 1 = 2008:2011; cycle 2 = 2012:2015; cycle 3 = 2016:2019. Macrogroups are sorted from high to low latitude.
Figure 12. Trends in sapling basal area (BA; sq.m/ha) by species group by Macrogroup and cycle. Saplings are stems that are >= 1 cm and < 10 cm DBH. Cycle 1 = 2008:2011; cycle 2 = 2012:2015; cycle 3 = 2016:2019. Macrogroups are sorted from high to low latitude.
Figure 13. Trends in sapling density (stems/sq.m) by species group by Macrogroup and cycle. Saplings are stems that are >= 1 cm and < 10 cm DBH. Cycle 1 = 2008:2011; cycle 2 = 2012:2015; cycle 3 = 2016:2019. Macrogroups are sorted from high to low latitude.
Figure 14. Trends in seedling density (stems/sq.m) by species group by Macrogroup and cycle. Seedlings are stems that are >= 15 cm tall and < 1 cm DBH. Cycle 1 = 2008:2011; cycle 2 = 2012:2015; cycle 3 = 2016:2019. Macrogroups are sorted from high to low latitude.
Figure 15. Live tree diameter at breast height (DBH) distribution in 10 cm increments by Macrogroup and cycle. Trees are stems that are >=10 cm DBH. Cycle 1 = 2008:2011; cycle 2 = 2012:2015; cycle 3 = 2016:2019. Macrogroups are sorted from high to low latitude.
Figure 16. Trends in live tree basal area (BA; sq.m/ha) for trees 10-20 cm DBH by Macrogroup and cycle. Cycle 1 = 2008:2011; cycle 2 = 2012:2015; cycle 3 = 2016:2019. Macrogroups are sorted from high to low latitude.
Figure 17. Trends in live tree basal area (BA; sq.m/ha) for trees 20-30 cm DBH by Macrogroup and cycle. Cycle 1 = 2008:2011; cycle 2 = 2012:2015; cycle 3 = 2016:2019. Macrogroups are sorted from high to low latitude.
Figure 18. Trends in live tree basal area (BA; sq.m/ha) for trees 30-40 cm DBH by Macrogroup and cycle. Cycle 1 = 2008:2011; cycle 2 = 2012:2015; cycle 3 = 2016:2019. Macrogroups are sorted from high to low latitude.
Figure 19. Trends in live tree density (stems/ha) for trees 10-20 cm DBH by Macrogroup and cycle. Cycle 1 = 2008:2011; cycle 2 = 2012:2015; cycle 3 = 2016:2019. Macrogroups are sorted from high to low latitude.
Figure 20. Trends in live tree density (stems/ha) for trees 20-30 cm DBH by Macrogroup and cycle. Cycle 1 = 2008:2011; cycle 2 = 2012:2015; cycle 3 = 2016:2019. Macrogroups are sorted from high to low latitude.
Figure 21. Trends in live tree density (stems/ha) for trees 30-40 cm DBH by Macrogroup and cycle. Cycle 1 = 2008:2011; cycle 2 = 2012:2015; cycle 3 = 2016:2019. Macrogroups are sorted from high to low latitude.
Figure 22. Trends in Sorensen similarity between sapling and canopy species by Macrogroup and cycle. Sorensen similarity is presence-only based similarity that ranges from 0 to 1. A score of 0 indicates there are no species in common between the two strata. A score of 1 indicates all species are in common between strata. Cycle 1 = 2008:2011; cycle 2 = 2012:2015; cycle 3 = 2016:2019. Macrogroups are sorted from high to low latitude.
Figure 23. Trends in Sorensen similarity between seedling and canopy species by Macrogroup and cycle. Sorensen similarity is presence-only based similarity that ranges from 0 to 1. A score of 0 indicates there are no species in common between the two strata. A score of 1 indicates all species are in common between strata. Cycle 1 = 2008:2011; cycle 2 = 2012:2015; cycle 3 = 2016:2019. Macrogroups are sorted from high to low latitude.
Figure 24. Percent of sufficiently stocked plots in most recent 4-year period (2016:2019), sorted from high to low by Macrogroup. Parks without bars had no sufficiently stocked plots. The level required for a plot to be stocked was adjusted deer browse impact, where parks averaging > 3 (moderate) for deer browse impact required a stocking index of 100 or more. Macrogroups averaging 3 or less for deer browse impact required a stocking index of 50 or more. The dashed line indicates the threshold used to determine critically low stocking in Figure 2.
Figure 25. Average stocking index in most recent 4-year period (2016:2019), sorted from high to low by Macrogroup. Stocking index only includes native canopy-forming species. Note that ash species (Fraxinus spp.) were not included in the index. Error bars are bootstrapped 95% confidence intervals.
Figure 26. Proportion of suboptimal species groups comprising regeneration layers in recent 4-year period (2016:2019). Macrogroups are sorted from high to low latitude.
Figure 27. Proportion of suboptimal species comprising regeneration layers in recent 4-year period (2016:2019). Macrogroups are sorted from high to low latitude.
Figure 28. Sapling density of native canopy-forming species in most recent 4-year period (2016:2019), sorted from high to low by Macrogroup. Note that ash species (Fraxinus spp.) were not included as canopy species. Error bars are bootstrapped 95% confidence intervals.
Figure 29. Seedling density of native canopy-forming species in most recent 4-year period (2016:2019), sorted from high to low by Macrogroup. Note that ash species (Fraxinus spp.) were not included as canopy species. Error bars are bootstrapped 95% confidence intervals.
Figure 30. Sorensen similarity between sapling and canopy species in most recent 4-year period (2016:2019), sorted from high to low by Macrogroup. Sorensen similarity is presence-only based similarity that ranges from 0 to 1. A score of 0 indicates there are no species in common between the two strata. A score of 1 indicates all species are in common between strata. Error bars are bootstrapped 95% confidence intervals.
Figure 31. Sorensen similarity between seedling and canopy species in most recent 4-year period (2016:2019), sorted from high to low by Macrogroup. Sorensen similarity is presence-only based similarity that ranges from 0 to 1. A score of 0 indicates there are no species in common between the two strata. A score of 1 indicates all species are in common between strata. Error bars are bootstrapped 95% confidence intervals.